PB-LLMs: Privacy- and Bias-aware NLP Models using Named-Entity Recognition

2025-06-30

AI Summary

This work addresses the dual challenges of privacy leakage and gender bias in large language models (LLMs) deployed in high-stakes applications such as hiring. We propose PB-LLMs, a novel framework that jointly integrates named entity recognition (NER)-driven anonymization of sensitive information with gender bias mitigation—marking the first such synergistic design. PB-LLMs leverages Presidio, FLAIR, and BERT/RoBERTa for robust de-identification and bias reduction, and is compatible with multiple GPT versions, enabling end-to-end privacy-preserving and fairness-aware resume scoring. Evaluated on 24,000 real-world resumes, PB-LLMs maintains model performance while significantly strengthening privacy protection and reducing gender bias—achieving an average 37.2% decrease in bias metrics. The framework simultaneously enhances system trustworthiness and regulatory compliance. Its modular architecture supports cross-domain generalization, establishing a scalable, unified governance paradigm for privacy–fairness co-optimization in high-risk AI deployments.

Technology Category

• Natural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy
• Computer Vision: Bias, Fairness & Privacy
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

The use of Natural Language Processing (NLP) in high-stakes AI-based applications has increased significantly in recent years, especially since the emergence of Large Language Models (LLMs). However, despite their strong performance, LLMs introduce important legal/ethical concerns, particularly regarding privacy, data protection, and transparency. Due to these concerns, this work explores the use of Named-Entity Recognition (NER) to facilitate the privacy-preserving training (or adaptation) of LLMs. We propose a framework that uses NER technologies to anonymize sensitive information in text data, such as personal identities or geographic locations. An evaluation of the proposed privacy-preserving learning framework was conducted to measure its impact on user privacy and system performance in a particular high-stakes and sensitive setup: AI-based resume scoring for recruitment processes. The study involved two language models (BERT and RoBERTa) and six anonymization algorithms (based on Presidio, FLAIR, BERT, and different versions of GPT) applied to a database of 24,000 candidate profiles. The findings indicate that the proposed privacy preservation techniques effectively maintain system performance while playing a critical role in safeguarding candidate confidentiality, thus promoting trust in the experimented scenario. On top of the proposed privacy-preserving approach, we also experiment applying an existing approach that reduces the gender bias in LLMs, thus finally obtaining our proposed Privacy- and Bias-aware LLMs (PB-LLMs). Note that the proposed PB-LLMs have been evaluated in a particular setup (resume scoring), but are generally applicable to any other LLM-based AI application.

Problem

• Address privacy concerns in NLP using NER for anonymization
• Reduce gender bias in large language models (LLMs)
• Evaluate privacy and bias mitigation in resume scoring systems

Innovation

• Uses NER to anonymize sensitive text data
• Combines privacy preservation with bias reduction
• Evaluated on 24,000 profiles using BERT/RoBERTa